Hacker News new | past | comments | ask | show | jobs | submit login
The Sad End of Intel Desktop Boards (os2museum.com)
117 points by eaguyhn on Nov 13, 2019 | hide | past | favorite | 61 comments

Now that Intel exited the business, what is the motherboard brand for people who just want a motherboard that works? One without any lights on it at all. It seems like supermicro would be the straight-faced alternative, but they don’t pay attention and often their boards just don’t work. I want a board that advertises its attention to signal integrity and thermal safety.

It seems backwards but late revision high end gaming boards have economies of scale that force the manufacturers into making stable boards to reduce RMA volume. The boards also tend to be cheaper too due the same economies of scale.

The VRM and other power delivery circuits on overclocking boards are also typically engineered to handle much higher voltages than spec.

Avoid the first revision though, which is probably good advice for any device.

Do the gaming boards support ECC memory however? That's usually the one major difference between a high performance gaming board and a basic workstation board. Sometimes they'll even have a chipset that has ECC support, but the BIOS will disable it because they hate you.

That's the big question. It seems like many AMD AM4 socket boards claim to support it. Few Intel boards do, unless you go with a Xeon socket. You can never really tell if they've done it properly until you get one on the bench and decode its DMI tables.

Although they certain have boards with flashy lights, Gigabyte has been my mainstay for stable boards, and they do advertise signal integrity and thermal safety. (They also have a Black Edition of many of their boards, which has extra testing)

Side note, for more capable boards with the features I need I've noticed it actually costs more to get one without built-in RGB. Not sure how that became the premium option now.

The lights are just personally offensive to me. I'm happy to pay more for their non-existence.

Guessing because desoldering is so time-consuming.

I've always had really good results (performance and longevity) with Asus gaming boards despite them having all the LED "features." I just turn all that junk off in the bios.

Be careful with Asus motherboards: it seems that Z390 boards will push some Asus software to Windows installations just by booting it unless you disable it properly in the BIOS.

As a consequence, you probably want to disable the BIOS option before booting any Windows installation (in my case, it was called "Q-installer", but it is different for gaming boards in my understanding). Until you disable it, it will reinstall and re-enable itself at boot.

See: https://www.techpowerup.com/248827/asus-z390-motherboards-au...

In my case, I noticed it thanks to Windows Server pointing out which service was slowing down shutdown, which was "AsusUpdateCheck". Increased the shutdown time by a few seconds.

I did not find any info about how to make Windows ignore the WPBT table.

Anecdotal, but my Asus Z170i from 2015 has been solid for 4 years. Doesn't have any lights, maybe there wasn't room for them in the miniITX form factor.

EDIT - had the date wrong, it was mid-2016.

Also anecdotal:

My desktop is still running smoothly on its Asus mobo from 2014 (I believe it was the Z97-P but I'm not home at the moment to double check). No fancy RGB LEDs, but I haven't shopped for one in a while so that may be harder to accomplish now. Not sure.

Either way, they've always been competitively priced and well supported. Used them to build both gaming and general use machines for friends/family over the past several years without issue.

Supermicro for fully fledged workstation experience

ASUS as a good "home" option

In reality I would suggest investing in HP Z series workstations or Z series small formfactor machines. They rock!

When I started building my own machines I originally went the inexpensive route with AMD CPUs and (I forget the brand) boards, but after three motherboard failures in 18 months I switched to Intel CPUs and Intel boards and never had a problem again. I haven't built a machine in years (I buy laptops, Mac minis, and Intel NUCs for my computing needs) and while it's a little sad that Intel is out of this game now if I was going to build out a Windows/Linux/*BSD non-laptop system it would likely be a NUC anyway so, in a way, I don't feel like I would be missing anything.

FYI, For the past year and a half (since the 2000 series Ryzen CPUs) the upper-mid range motherboards have MUCH better designs in terms of power and thermal management than a few years ago. With Ryzen's relative popularity with the home/enthusiast builders, it's become worth it for mfgs to prioritize those efforts more.

You should have a better experience today. Though, I would wait 2-3 months after a product (mb/cpu) launch before getting one in general. Just so BIOS updates can shake out a bit and software support (Windows/Linux) can as well.

I used to buy Intel boards because they were more likely to work well with the latest CPUs of the day. I've switched to SuperMicro boards for our desktops (which are mostly dual Xeon boxes) and have had few problems.

I wish they'd come with a TPM. The integration issues are frequently around the TPM and getting it recognized. It's required for BitLocker whole disk encryption which is part of our standard build. (If a machine or drive goes AWOL, we don't have to worry.)

I used to go for boards with a physical TPM as well(the Qxxx) and had the same driver issues you mentioned. But I've been reading recently that it might not be needed, since some of the newer intel chipsets(like z390) now have builtin Intel® Platform Trust Technology (Intel® PTT) which supposedly is a sort of firmware based TPM. I plan to give it a shot and see if it works for bitlocker hardware encryption when I buy my next motherboard.

In case it's helpful for you to know, you can set up Bitlocker without a TPM. Google for instructions. It works well.

Don't you need to type in an additional password every time you boot?

I have an intel board in my current daily desktop PC, it's socket LGA1150 also, but previous generation, Ivy Bridge. Of all the desktop motherboards I've had (and I've gone through a lot) it's the most stable I've ever used.

I worked for a small company who built white-box PCs in the late 90's thru the early 2000's. The Intel desktop boards were far and away more reliable than any other manufacturer's boards. The Intel boards weren't flashy or as feature-filled, and they did have some "duds" (particulat board models with various "erratum", the whole 820 chipset SDRAM controller debocle), but their manufacturing consistency was excellent. I don't believe I ever encountered a DOA board coming out of the box, which I can't say for ASUS or Gigabyte.

Edit: Power supply quality helps a lot, too, as others have said. I can't remember the brand of power supplies we used, but that definitely mattered.

They've also been faster at removing older tech, and getting newer stuff integrated when they were building/designing them. I remember getting one of the first Intel boards without a floppy controller (or parallel/serial headers). Of course bit me in the ass trying to get Windows Server on the thing since the extra device drivers for HDD controller was via floppy.

Power supply issues have bitten me in the past probably more than anything else in terms of issues.

Same. I built a modest core i3 machine in like 2012 for my dad. That computer has had zero problems and runs Windows 7 with uptimes comparable to Linux servers. I never really thought about it until now, but the Intel motherboard might be the reason.

Not really, that is the norm for most decent brands.

Bad PSUs are probably a bigger factor.

I had to upgrade and couldn't find an Intel board. Choose from what was available, not cheap, and suddenly my PC has a personality (in a bad way).

Are you overclocking or doing other fancy things? Last time I had an issue with a board was core 2 duo times. PCIe would magically cripple itself to 1x for all slots as soon as I overclocked the CPU. There was a workaround for that on the gigabyte forums. But that's about every problem I ever had.

Overclocking desktops in 2019 is only extremely rarely a productive use of your time.

Overclocking desktop in 2019 is a 10 minute affair to gain a permanent >20% performance.

I think you're saying the same thing. A one-time 20% bump isn't worth that much these days to the vast majority of people. (Even if it is only 10 minutes... which I tend to think might be an underestimate, given the testing, etc. you'd have to do to make sure it's all stable. 10 minutes to an overclocked boot, fine. 10 minutes to an overclocked configuration that can run a workload without problems for eight hours a day for several years... that seems like more of a challenge).


That's all true (I personally don't OC either), but you should still run stability checks when you first get your computer. This Coding Horror post is great:


IMO the use of Linux for out of the box stability testing is poor unless you're absolutely certain the hardware config works perfectly under Linux without any specific tinkering.

Pretty much any PC will boot linux nowadays, maybe you need wifi or graphics drivers, but that post only uses Linux to test the CPU and disk.

I think it's a little less clear of a performance benefit in 2019, now that every CPU will dynamically overclock itself within thermal limits.

Intel parts only dynamically OC to their max turbo presets. you can often go quite a bit further if you don't care about idle power use. I was able to OC my 9700k to run 5 GHz on all cores and tighten up the AVX offset to -1 (I think it's -3 stock).

can I notice the difference outside of benchmarks? probably not, but the performance increase per core is comparable to stepping up to a stock 9700k from a 9600k (ignoring core count). if you would pay $100 to have a 9700k instead of a 9600k (which I did), why not overclock it too? it took me about 20 minutes of actual work and 12 hours of waiting for stress tests to fail.

Honestly I don't think so. I tried overclocking my i7-4790k back in the day and it took an absurd amount of time and I never really got a meaningful improvement that I could honestly say was stable.

I hear the i9-9900k is a good overclocker but with AMD being the best option now and their chips being well, let's be honest, not the best overclockers in the world I don't really see the point.

Which current gen CPUs can you get >20% performance increase on by overclocking? Assuming that baseline performance is stock settings with turbo enabled (usually default) as well as MCE/PBO (default on some motherboards).

Turns out that I'd rather have longer life and lower power draw over 20% performance most of the time.

It's well worth it, especially when it's on a triple radiator water cooling loop

I used to own an Intel Ivy. It simply wouldn't boot at all with the smallest possible OC (yes, I had a 'K) and would automatically revert.

Never had stability problems though.

I did not know Intel stayed in the desktop motherboard business as long as they did. I had an AL440LX motherboard-based system that I'm sure still works, after a few battery and power supply replacements, if it hasn't been disposed of. (I'm sure it has been disposed of)

The author fails to understand some things:

Not so for the owners of Intel boards. To show just much Intel values its customers, they were informed that BIOS updates were not forthcoming and the newer, faster processors were not supported on Intel boards. That was no doubt particularly galling to the owners of the DZ87KLT boards, which sold for around $300 when new and are worth $150 or more even today (2019).


After all the hoopla in 2018 about Meltdown and Spectre, guess what happened: Intel somehow magically managed to update the BIOS for those boards after akl. For the DZ87KLT-75K, nothing really changed because the ME firmware did not get updated. But for the DQ87PG board, the 2018 BIOS updates did update the ME firmware as well.

It's precisely because Intel "values its customers," or more precisely because Intel understands its value add for customers versus the other motherboard manufacturers, that they would not just slide support for a few extra CPUs into the firmware. To do the job to their own standards they would need to fully retest everything with those new CPUs before officially supporting them. The economics of the Meltdown and Spectre fixes is an entirely separate matter.

(it'd be fair to point out that home power users who want to upgrade the CPU on a motherboard after purchase were never a large proportion of the Intel motherboard buying public and, in recent years... there cannot be that many)

Ha. I recently upgraded to a 4790s and a intel desktop board. I thought the restarting problem was a faulty board, now I know better! Wonder how hard it is to update the ME...when I updated the bios the problem was still there.

The intel drivers page for your board should have an update package for the ME (if one is available).

Following a few links and searches there is a github page with all the ME bins and related tools. The key is installing a newer version of ME than comes with the board BIOS.

Looking at the attitude of locked down chipsets and ME, would you expect any better? Intel's goal is to sell a product that fulfills some short-sighted requirements and never look back.

Their main business customers aren't interested in mixing and matching pieces of already purchased computers. Just like they don't particularly care that the whole house of cards is theoretically subservient to some inauditable backdoor.

That's how I feel about the NUCs. They are only being produced by Intel to try and advance a market.

Intel makes processors. The will sell things like motherboards, Compute Sticks, and NUCs just to push their primary line. Even their Flash, Intel doesn't want to be a Flash Fab, but it helps their processor line in many ways to have a presence.

It's like the Microsoft Surface. Microsoft will drop support for it in a hot minute when they feel the Surface has succeeded or failed. They already did it once with the Surface RT, but maybe I'm wrong and Microsoft is pivoting with the Surface to be more Apple-like... lol

What does the article mean by the "end" of desktop boards? Is it the end of sockets accepting multiple generations of CPUs?

I can see that it's not entirely true either, Socket 1151 supports both Skylake and Kaby Lake: http://asrock.com/mb/Intel/Z390%20Phantom%20Gaming%20X/index...

What change did actually take place?

They used to make their own motherboards and sell them to consumers. They were known for being very high quality and very reliable. Now you can only get boards for their chips from third party OEMs who tend to have much more variable quality (which vendors are "best" tends to change each board/chipset generation).

I dunno about that. If you're buying whatever's on sale, yeah, you're at the whims of the stamp-it-out gods. But I build a lot of PCs and Gigabyte, for example, has been consistently reliable across...five or six generations of Intel chips? Plus some AMD builds in there. Absolutely zero failures.

Anecdotal, but it also seems to line up with most of what I hear out there.

I see a lot of mentions of Gigabyte and Asus in this thread -- I have had good luck with MSI boards too.

Maybe the reason Intel quit making boards is because there are lots of good motherboards now?

It might also matter what "tier" of board you buy.

Getting the "basement bargain" board for $20... well, you get what you pay for.

But for each of those top brands, Gigabyte, ASUS, and MSI, buying their top tier products gets you a seriously good board.

The gaming boards tend to have a lot of features a workstation user might not need, like great overclocking support, etc... but with that quality comes stability. Worth every penny to get a good motherboard when doing a build.

That... and I no longer get cheap cases either... but that's another story. The moral? Don't skimp on your components, and you'll wind up with a much better system that lasts you for years.

With cases, I think it's less about price and more about good airflow design, which will often mean buying more fans and/or a fan hub. There are really good cases from as low as $40 that will do better than some $120+ cases.

The problem with cheap cases does tend to be airflow and locations for fans... but also the build quality of the case.

Things like ports for routing cables without causing a rat's nest... cheap stamped sheet metal with sharp edges to cut yourself on, decent I/O backplate, temperature control zones, bottom mounted PSU, etc. A good quality case will cost you... but you can likely get several builds out of it over the years... so well worth it.

While I agree... performance is also an issue... Personally, I've been using the various Fractal Define series cases more often than not. Latest personal build is a Lian Li O11 Dynamic Razor Edition (frankly regret going RGB on a Linux box).

I'd agree with this. I've never bought a "high end" board, either. Even the middle ground is fine.

Had an unstable Asus board. We bought 3 in the company, got them all replaced - I think we got a slightly newer revision back. Same problem, watch video for 5-10 minutes and it would crash, despite temperature being fine.

Eventually I moved the processor to an Intel board (so yes, this is some years back, probably just before they stopped making them), and had no instability. The Asus motherboard was somehow incompatible with the processor, I guess.

So I didn't go with Asus last time I bought a motherboard, although I wouldn't be surprised if they're otherwise fine.

Even different boards with the same chipset from the same mfg will vary a lot.

Intel used to sell motherboards too which anecdotally I seem to recall them being expensive but performant and this article covers how around 2013 when Intel got out of the motherboard business their support of next gen processors (i.e. post 2013) on existing motherboards was lackluster.

I wish Intel would allow third party chipsets again. I liked the days when companies would compete on features like more PCI lanes and faster interconnects.

The PCIe root ports are integrated with the CPUs these days. If you want more slots, get a breakout box. Nobody can get you more lanes.

Or get just AMD.

It is not clear the situation, but guess may be you need to have those board.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact